>: Contrary to what the FAQ says, Linux can run on MCA, and has been for
>: about two or three years. It just hasn't been integrated into the
>: standard kernel. Why? Beats me.
>Actually some of the earliest Linux kernels will run on MCA machines,
>just that they cannot access any devices on the MCA bus (such as HDD's)
Some MCA machines can run some Linux kernels. The model 55SX
with a Adaptec 1640 SCSI controller, for example, will run a SCSI
kernel. Unfortunately, most other models require that IRQ 0 be reset
after a timer interrupt because of level triggered interrupts. Since the
standard Linux kernel doesn't do this, these machines will hang after the
first timer interrupt.
Anyhow, saying that you can run the kernel without any devices is about
the same as saying you can't run the kernel. I personally don't have
much use for an OS that can just run an idle task.
>: What amuses me the most is people who _insist_ that Linux can't run on
>: MCA, and never can because it's a different architecture. One could say
>The code to handle MCA is most likely simpler than that for PCI.
If you were to use _all_ the various MCA features, I think it could get
as complex as PCI. The adapter detection code can be as complex, and
then you can have bus error handlers, shared IRQs, etc. I don't know
how bus extensions would be dealt with. Plus, in theory you could
build a MCA SMP system by just adding a CPU board to the bus, which
adds a whole new level of complexity. On the other hand, simple
adapter detection and shared IRQs doesn't even come close to pci.c in
size or complexity, and this is more than sufficient to get it up and
running (actually, you could ditch shared IRQs, but why would you want